Skip to main content

Table 1 Graph characteristics for the analysed datasets. Here, ‘node degree’ refers to the count of all adjacent edges to a given node (i.e. in-degree + out-degree)

From: Assessing the effects of hyperparameters on knowledge graph embedding quality

Statistic

FB15k-237

UMLS

WN18RR

node count

14,541

135

40,943

edge count

310,116

6529

93,003

edge type count

237

46

11

density

6.19e−06

7.85e−03

5.04e−06

component count

6

1

13

mean component diameter

2.5

2

2.46

mean component distance

1.40

1.61

1.55

mean component connectivity

1

4

1.15

mean node degree

42.65

96.73

4.54

median node degree

26

71

3

maximum node degree

8642

382

521

standard deviation of node degree

127.70

87.44

8.58

skewness of node degree

34.07

1.84

26.15

kurtosis of node degree

1777.59

2.91

1095.90

  1. A ‘component’ refers to a weakly connected subgraph that is disconnected from the rest of the graph